Ray Traversal Research Articles

Locally-Adaptive Level-of-Detail for Hardware-Accelerated Ray Tracing

We introduce an adaptive level-of-detail technique for ray tracing triangle meshes that aims to reduce the memory bandwidth used during ray traversal, which can be the bottleneck for rendering time with large scenes and the primary consumer of energy. We propose a specific data structure for hierarchically representing triangle meshes, allowing localized decisions for the desired mesh resolution per ray. Starting with the lowest-resolution triangle mesh level, higher-resolution levels are generated by tessellating each triangle into four via splitting its edges with arbitrarily-placed vertices. We fit the resulting mesh hierarchy into a specialized acceleration structure to perform on-the-fly tessellation level selection during ray traversal. Our structure reduces both storage cost and data movement during rendering, which are the main consumers of energy. It also allows continuous transitions between detail levels, while locally adjusting the mesh resolution per ray and preserving watertightness. We present how this structure can be used with both primary and secondary rays for reflections and shadows, which can intersect with different tessellation levels, providing consistent results. We also propose specific hardware units to cover the cost of additional compute needed for level-of-detail operations. We evaluate our method using a cycle-accurate simulation of a custom ray tracing hardware architecture. Our results show that, as compared to traditional bounding volume hierarchies, our method can provide more than an order of magnitude reduction in energy use and render time, given sufficient computational resources.

A parallel construction method based on linear hierarchical bounding box without stack

The intersect operation of geometry is a common method of collision detection. To achieve a fast intersect operation, a hierarchical bounding box must be quickly constructed. This paper proposes an effective space division method for constructing high-quality linear hierarchical bounding boxes through ray tracing. This method can generate more regular axial bounding boxes into a complete binary tree. This structure can be completely parallel on the GPU, and effective parallelization can ensure the fastest tree-building time. The algorithm in this paper uses a non-stack ray tracing method to increase the frame rate of the 3D model rendering in the graphics processing unit and reduce the construction time of the bounding box. The experimental analysis of construction time, frame rate, average number of intersections and other performance indicators, the experimental results show that the use of stack-free midpoint splitting method and surface heuristic splitting method to ensure rapid construction time and efficient ray traversal performance, improve the object’s light traversal efficiency.

Data-driven dose calculation algorithm based on deep U-Net

Accurate and efficient dose calculation is an important prerequisite to ensure the success of radiation therapy. However, all the dose calculation algorithms commonly used in current clinical practice have to compromise between calculation accuracy and efficiency, which may result in unsatisfactory dose accuracy or highly intensive computation time in many clinical situations. The purpose of this work is to develop a novel dose calculation algorithm based on the deep learning method for radiation therapy. In this study we performed a feasibility investigation on implementing a fast and accurate dose calculation based on a deep learning technique. A two-dimensional (2D) fluence map was first converted into a three-dimensional (3D) volume using ray traversal algorithm. 3D U-Net like deep residual network was then established to learn a mapping between this converted 3D volume, CT and 3D dose distribution. Therefore an indirect relationship was built between a fluence map and its corresponding 3D dose distribution without using significantly complex neural networks. Two hundred patients, including nasopharyngeal, lung, rectum and breast cancer cases, were collected and applied to train the proposed network. Additional 47 patients were randomly selected to evaluate the accuracy of the proposed method through comparing dose distributions, dose volume histograms and clinical indices with the results from a treatment planning system (TPS), which was used as the ground truth in this study. The proposed deep learning based dose calculation algorithm achieved good predictive performance. For 47 tested patients, the average per-voxel bias of the deep learning calculated value and standard deviation (normalized to the prescription), relative to the TPS calculation, is 0.17%±2.28%. The average deep learning calculated values and standard deviations for relevant clinical indices were compared with the TPS calculated results and the t-test p-values demonstrated the consistency between them. In this study we developed a new deep learning based dose calculation method. This approach was evaluated by the clinical cases with different sites. Our results demonstrated its feasibility and reliability and indicated its great potential to improve the efficiency and accuracy of radiation dose calculation for different treatment modalities.

Hardware-Accelerated Dual-Split Trees

Bounding volume hierarchies (BVH) are the most widely used acceleration structures for ray tracing due to their high construction and traversal performance. However, the bounding planes shared between parent and children bounding boxes is an inherent storage redundancy that limits further improvement in performance due to the memory cost of reading these redundant planes. Dual-split trees can create identical space partitioning as BVHs, but in a compact form using less memory by eliminating the redundancies of the BVH structure representation. This reduction in memory storage and data movement translates to faster ray traversal and better energy efficiency. Yet, the performance benefits of dual-split trees are undermined by the processing required to extract the necessary information from their compact representation. This involves bit manipulations and branching instructions which are inefficient in software. We introduce hardware acceleration for dual-split trees and show that the performance advantages over BVHs are emphasized in a hardware ray tracing context that can take advantage of such acceleration. We provide details on how the operations needed for decoding dual-split tree nodes can be implemented in hardware and present experiments in a number of scenes with different sizes using path tracing. In our experiments, we have observed up to 31% reduction in render time and 38% energy saving using dual-split trees as compared to binary BVHs representing identical space partitioning.

Rasterisation-based progressive photon mapping

Ray tracing on the GPU has been synergistically operating alongside rasterisation in interactive rendering engines for some time now, in order to accurately capture certain illumination effects. In the same spirit, in this paper, we propose an implementation of progressive photon mapping entirely on the rasterisation pipeline, which is agnostic to the specific GPU architecture, in order to synthesise images at interactive rates. While any GPU ray tracing architecture can be used for photon mapping, performing ray traversal in image space minimises acceleration data structure construction time and supports arbitrarily complex and fully dynamic geometry. Furthermore, this strategy maximises data structure reuse by encompassing rasterisation, ray tracing and photon gathering tasks in a single data structure. Both eye and light paths of arbitrary depth are traced on multi-view deep G-buffers, and photon flux is gathered by a properly adapted multi-view photon splatting. In contrast to previous methods exploiting rasterisation to some extent, due to our novel indirect photon splatting approach, any event combination present in photon mapping is captured. We evaluate our method using typical test scenes and scenarios for photon mapping methods and show how our approach outperforms typical GPU-based progressive photon mapping.

Efficient ray casting of volumetric images using distance maps for empty space skipping

Volume and isosurface rendering are methods of projecting volumetric images to two dimensions for visualisation. These methods are common in medical imaging and scientific visualisation.Head-mounted optical see-through displays have recently become an affordable technology and are a promising platform for volumetric image visualisation. Images displayed on a head-mounted display must be presented at a high frame rate and with low latency to compensate for head motion. High latency can be jarring and may cause cybersickness which has similar symptoms to motion sickness.Volumetric images can be very computationally expensive to render as they often have hundreds of millions of scalar values. Fortunately, certain materials in images such as air surrounding an object boundary are often made transparent and need not be sampled, which improves rendering efficiency.In our previous work we introduced a novel ray traversal technique for rendering large sparse volumetric images at high frame rates. The method relied on the computation of an occupancy and distance map to speed up ray traversal through empty regions.In this work we achieve higher frame rates than our previous work with an improved method of resuming empty space skipping and the use of anisotropic Chebyshev distance maps. An optimised algorithm for computing Chebyshev distance maps on a graphical processing unit is introduced supporting real-time transfer function editing.

Parallel Reinsertion for Bounding Volume Hierarchy Optimization

AbstractWe present a novel highly parallel method for optimizing bounding volume hierarchies (BVH) targeting contemporary GPU architectures. The core of our method is based on the insertion‐based BVH optimization that is known to achieve excellent results in terms of the SAH cost. The original algorithm is, however, inherently sequential: no efficient parallel version of the method exists, which limits its practical utility. We reformulate the algorithm while exploiting the observation that there is no need to remove the nodes from the BVH prior to finding their optimized positions in the tree. We can search for the optimized positions for all nodes in parallel while simultaneously tracking the corresponding SAH cost reduction. We update in parallel all nodes for which better position was found while efficiently handling potential conflicts during these updates. We implemented our algorithm in CUDA and evaluated the resulting BVH in the context of the GPU ray tracing. The results indicate that the method is able to achieve the best ray traversal performance among the state of the art GPU‐based BVH construction methods.

RaTrace: simple and efficient abstractions for BVH ray traversal algorithms

In order to achieve the highest possible performance, the ray traversal and intersection routines at the core of every high-performance ray tracer are usually hand-coded, heavily optimized, and implemented separately for each hardware platform—even though they share most of their algorithmic core. The results are implementations that heavily mix algorithmic aspects with hardware and implementation details, making the code non-portable and difficult to change and maintain. In this paper, we present a new approach that offers the ability to define in a functional language a set of conceptual, high-level language abstractions that are optimized away by a special compiler in order to maximize performance. Using this abstraction mechanism we separate a generic ray traversal and intersection algorithm from its low-level aspects that are specific to the target hardware. We demonstrate that our code is not only significantly more flexible, simpler to write, and more concise but also that the compiled results perform as well as state-of-the-art implementations on any of the tested CPU and GPU platforms.

Ray tracing method for the description of radiation trapping in 3D plasma domains

A new approach for the solution of the Holstein–Biberman equation based on the advanced matrix method is developed. It allows for the consideration of radiation trapping in arbitrary finite 3D plasma domains for the various shapes of line contours and in a wide range of optical depths. Homogeneous and inhomogeneous distributions of absorbing atoms are considered. To solve the equation, an arbitrary plasma domain is discretized on a Cartesian voxel grid. The distances between the cells which are crossed by photons are computed by means of an efficient ray traversal algorithm. The algorithm is optimized for parallel computation on a graphical processing unit (GPU). For the Lorentzian shape of emission and absorption lines, the analytical expressions (which significantly decrease the computation time) have been derived. In the high opacity limit, the matrix is transformed to the universal form with an escape factor as a multiplier. The method is validated against a previously developed matrix approach by comparing the solutions for a finite cylinder geometry. The applicability range of the old method is specified. This range is defined by the asymptotics of Lorentz line wings at high optical depths. The capability of the method is illustrated with several complex geometries which are typical for various plasma sources. The effects related to the presence of photon blocking barriers are demonstrated. The proposed method allows for the demonstration of the fundamental differences between radiation and diffusion transport processes in the plasma domains of a complex shape. The method can be integrated into multi-component collisional–radiative models.

GPU Ray Tracing using Irregular Grids

We present a spatial index structure to accelerate ray tracing on GPUs. It is a flat, non-hierarchical spatial subdivision of the scene into axis aligned cells of varying size. In order to construct it, we first nest an octree into each cell of a uniform grid. We then apply two optimization passes to increase ray traversal performance: First, we reduce the expected cost for ray traversal by merging cells together. This adapts the structure to complex primitive distributions, solving the teapot in a stadium problem. Second, we decouple the cell boundaries used during traversal for rays entering and exiting a given cell. This allows us to extend the exiting boundaries over adjacent cells that are either empty or do not contain additional primitives. Now, exiting rays can skip empty space and avoid repeating intersection tests. Finally, we demonstrate that in addition to the fast ray traversal performance, the structure can be rebuilt efficiently in parallel, allowing for ray tracing dynamic scenes.

Advanced Approach for Radiation Transport Description in 3D Collisional-radiative Models

The description of radiation transport phenomena in the frames of collisional-radiative models requires the solution of Holstein-Biberman equation. An advanced solutuion method for 3D plasma obejcts is proposed. The method is applicable for various line contours in a wide range of absorption coefficients. Developed approach is based on discretization of the arbitrary plasma volume on a Cartesian voxel grid. Transport of photons between the cells is computed using the ray traversal algorithm by Amanatides [1]. Solution of the particle balance equations with computed in advance radiative transfer matrix is demonstrated for various typical arc shapes, like e.g. free-burning arc and cylindric arc. Results are compared with corresponding calculations using previously developed approaches. As the method is suited for finite geometries and allows for a strict solution of the radiation transport equation, applicability ranges of previous approximations can be specified.

Parallel On-Demand Hierarchy Construction on Contemporary GPUs

We present the first parallel on-demand spatial hierarchy construction algorithm targeting ray tracing on many-core processors such as GPUs. The method performs simultaneous ray traversal and spatial hierarchy construction focused on the parts of the data structure being traversed. The method is based on a versatile framework built around a task pool and runs entirely on the GPU. We show that the on-demand construction can improve rendering times compared to full hierarchy construction. We evaluate our method on both object (BVH) and space (kd-tree) subdivision data structures and compare them mutually. The on-demand method is particularly beneficial for rendering large scenes with high occlusion. We also present SAH kd-tree builder that outperforms previous state-of-the-art builders running on the GPU.

Efficient construction of bounding volume hierarchies into a complete octree for ray tracing

AbstractThis paper proposes an efficient construction scheme for bounding volume hierarchies based on a complete tree. This construction offers up to 4× faster construction times than binned‐surface area heuristic and offers competitive ray traversal performance. The construction is fully parallelized on x86 CPU architectures; it takes advantage of the eight‐wide vector units and exploits the advance vector extensions available for current x86 CPU architectures. Additionally, this work presents a clustering algorithm for grouping primitives, which can be computed in linear time O(n). Furthermore, this construction uses the graphics processing unit to perform intensive operations efficiently. Copyright © 2016 John Wiley & Sons, Ltd.

Shell: A Spatial Decomposition Data Structure for Ray Traversal on GPU

Shared memory many-core processors such as GPUs have been extensively used in accelerating computation-intensive algorithms and applications. When porting existing algorithms from sequential or other parallel architecture models to shared memory many-core architectures, non-trivial modifications are often needed to match the execution patterns of the target algorithms with the characteristics of many-core architectures. Ray traversal is a fundamental process in many applications, and is commonly accelerated by spatial decomposition schemes captured in hierarchical data structures (e.g., kd-trees). However, ray traversal using hierarchical data structures needs to conduct repeated hierarchical searches. Such search process is time-consuming on shared memory many-core architectures since it incurs considerable amounts of expensive memory accesses and execution divergence. In this paper, we propose a novel spatial decomposition based data structure, called Shell, which completely avoids hierarchical search for ray traversal. In Shell, a structure is built on the boundary of each region in the decomposed space, which allows any ray traversing in a region to find the next neighboring region to traverse using table lookup schemes, without any hierarchical search. While our ray traversal approach works for other spatial decomposition paradigms and many-core processors in higher dimensional scenes, we illustrate it using kd-tree on GPU for 3D scenario and compare with the fastest known kd-tree searching algorithms for ray traversal. Experimental results in graphics ray tracing and radiation dose calculation show that our approach improves the performance by 3.5-5.5 $\times$ over the fastest known kd-tree based approaches.

HART: A Hybrid Architecture for Ray Tracing Animated Scenes.

We present a hybrid architecture, inspired by asynchronous BVH construction [1], for ray tracing animated scenes. Our hybrid architecture utilizes heterogeneous hardware resources: dedicated ray-tracing hardware for BVH updates and ray traversal and a CPU for BVH reconstruction. We also present a traversal scheme using a primitive's axis-aligned bounding box (PrimAABB). This scheme reduces ray-primitive intersection tests by reusing existing BVH traversal units and the primAABB data for tree updates; it enables the use of shallow trees to reduce tree build times, tree sizes, and bus bandwidth requirements. Furthermore, we present a cache scheme that exploits consecutive memory access by reusing data in an L1 cache block. We perform cycle-accurate simulations to verify our architecture, and the simulation results indicate that the proposed architecture can achieve real-time Whitted ray tracing animated scenes at 1,920 × 1,200 resolution. This result comes from our high-performance hardware architecture and minimized resource requirements for tree updates.

Study of a Ray Casting Technique for the Visualization of Deformable Volumes.

Deformable models are widely used in many disciplines such as engineering and medicine. Real objects are usually scanned to create models in such applications. In many cases the shape of the object is extracted from volumetric data acquired during the scanning phase. At the same time, this volume can be used to define the model's appearance. In order to achieve a visualization that unifies the shape (physical model) and appearance (scanned volume) specially adapted volume rendering techniques are required. One of the most common volumetric visualization techniques is ray casting, which also enables the use of different corrections or improvements such as adaptive sampling or stochastic jittering. This paper presents an extensive study about a ray casting method for tetrahedral meshes with an underlying structured volume. This allows a direct visualization of the deformed model without losing the information contained in the volume. The aim of this study is to analyse and compare the different methods for ray traversal and illumination correction, resulting in a comprehensive relation of the different methods, their computational cost and visual performance.

Dynamic ray stream traversal

While each new generation of processors gets larger caches and more compute power, external memory bandwidth capabilities increase at a much lower pace. Additionally, processors are equipped with wide vector units that require low instruction level divergence to be efficiently utilized. In order to exploit these trends for ray tracing, we present an alternative to traditional depth-first ray traversal that takes advantage of the available cache hierarchy, and provides high SIMD efficiency, while keeping memory bus traffic low. Our main contribution is an efficient algorithm for traversing large packets of rays against a bounding volume hierarchy in a way that groups coherent rays during traversal. In contrast to previous large packet traversal methods, our algorithm allows for individual traversal order for each ray, which is essential for efficient ray tracing. Ray tracing algorithms is a mature research field in computer graphics, and despite this, our new technique increases traversal performance by 36--53%, and is applicable to most ray tracers.

SATO: Surface Area Traversal Order for Shadow Ray Tracing

AbstractWe present the surface area traversal order (SATO) metric to accelerate shadow ray traversal. Our formulation uses the surface area of each child node to compute the TO. In this metric, we give a traversal priority to the child node with the larger surface area to quickly find occluders. Our algorithm reduces the pre‐processing overhead significantly, and is much faster than other metrics. Overall, the SATO is useful for ray tracing large and complex dynamic scenes (e.g. a few million triangles) with shadows.

Ray-Box Culling for Tree Structures *

Ray-primitive intersection tests are the most important operations in ray tracing. The ray-box culling algorithm was presented to accelerate these intersection tests on grid structures, but this algorithm has not been widely used due to the poor ray traversal performance of the grid structures themselves. In this paper, we demonstrate how to apply this algorithm to tree structures and investigate its efficiency in terms of tree construction time and ray traversal performance. Experimental results show that our approach achieves up to 1.15× faster ray tracing performance and up to 1.22× faster total rendering performance, including tree construction and ray tracing.

Octree Rasterization: Accelerating High-Quality Out-of-Core GPU Volume Rendering

We present a novel approach for GPU-based high-quality volume rendering of large out-of-core volume data. By focusing on the locations and costs of ray traversal, we are able to significantly reduce the rendering time over traditional algorithms. We store a volume in an octree (of bricks); in addition, every brick is further split into regular macrocells. Our solutions move the branch-intensive accelerating structure traversal out of the GPU raycasting loop and introduce an efficient empty-space culling method by rasterizing the proxy geometry of a view-dependent cut of the octree nodes. This rasterization pass can capture all of the bricks that the ray penetrates in a per-pixel list. Since the per-pixel list is captured in a front-to-back order, our raycasting pass needs only to cast rays inside the tighter ray segments. As a result, we achieve two levels of empty space skipping: the brick level and the macrocell level. During evaluation and testing, this technique achieved 2 to 4 times faster rendering speed than a current state-of-the-art algorithm across a variety of data sets.