An FPGA Acceleration for the Kd-tree Search in Photon Mapping

Kuhara, Takuya; Miyajima, Takaaki; Yoshimi, Masato; Amano, Hideharu

doi:10.1007/978-3-642-36812-7_3

Cited by 11 publications

(7 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Mesorasi maps N to the GPU because neighbor search is highly parallel, but does not map to the specialized datapath of an NPU. Alternatively, an SoC could use a dedicated neighbor search engine (NSE) [31], [59]. We use the GPU because it is prevalent in today's SoCs and thus provides a concrete context to describe our design.…”

Section: V-a Overall Designmentioning

confidence: 99%

See 1 more Smart Citation

Mesorasi: Architecture Support for Point Cloud Analytics via Delayed-Aggregation

Tian

et al. 2020

Preprint

View full text Add to dashboard Cite

Point cloud analytics is poised to become a key workload on battery-powered embedded and mobile platforms in a wide range of emerging application domains, such as autonomous driving, robotics, and augmented reality, where efficiency is paramount. This paper proposes Mesorasi, an algorithm-architecture co-designed system that simultaneously improves the performance and energy efficiency of point cloud analytics while retaining its accuracy.Our extensive characterizations of state-of-the-art point cloud algorithms show that, while structurally reminiscent of convolutional neural networks (CNNs), point cloud algorithms exhibit inherent compute and memory inefficiencies due to the unique characteristics of point cloud data. We propose delayed-aggregation, a new algorithmic primitive for building efficient point cloud algorithms. Delayed-aggregation hides the performance bottlenecks and reduces the compute and memory redundancies by exploiting the approximately distributive property of key operations in point cloud algorithms. Delayed-aggregation let point cloud algorithms achieve 1.6× speedup and 51.1% energy reduction on a mobile GPU while retaining the accuracy (-0.9% loss to 1.2% gains). To maximize the algorithmic benefits, we propose minor extensions to contemporary CNN accelerators, which can be integrated into a mobile Systems-on-a-Chip (SoC) without modifying other SoC components. With additional hardware support, Mesorasi achieves up to 3.6× speedup.

show abstract

Section: V-a Overall Designmentioning

confidence: 99%

“…Prior work has also extensively studied systems and architectures for accelerating neighbor search on GPU [25], [44], FPGA [27], [31], [56], and ASIC [59]. Neighbor search contributes non-trivial execution time to point cloud networks.…”

Section: Related Workmentioning

confidence: 99%

Mesorasi: Architecture Support for Point Cloud Analytics via Delayed-Aggregation

Tian

et al. 2020

Preprint

View full text Add to dashboard Cite

show abstract

“…Accelerating KD-tree search has been mostly explored in the context of Map-Reduce [5], GPU [23,28,36,53], and FPGA [71]. The Tigris accelerator differs from prior attempts in its systematic and comprehensive exploitation of different forms of parallelism in KD-tree search.…”

Section: Related Workmentioning

confidence: 99%

Tigris

Tian

Zhu

2019

Proceedings of the 52nd Annual IEEE/ACM International Symposium on Microarchitecture

View full text Add to dashboard Cite

Machine perception applications are increasingly moving toward manipulating and processing 3D point cloud. This paper focuses on point cloud registration, a key primitive of 3D data processing widely used in high-level tasks such as odometry, simultaneous localization and mapping, and 3D reconstruction. As these applications are routinely deployed in energy-constrained environments, real-time and energy-efficient point cloud registration is critical.We present Tigris, an algorithm-architecture co-designed system specialized for point cloud registration. Through an extensive exploration of the registration pipeline design space, we find that, while different design points make vastly different trade-offs between accuracy and performance, KD-tree search is a common performance bottleneck, and thus is an ideal candidate for architectural specialization. While KD-tree search is inherently sequential, we propose an acceleration-amenable data structure and search algorithm that exposes different forms of parallelism of KD-tree search in the context of point cloud registration. The co-designed accelerator systematically exploits the parallelism while incorporating a set of architectural techniques that further improve the accelerator efficiency. Overall, Tigris achieves 77.2× speedup and 7.4× power reduction in KD-tree search over an RTX 2080 Ti GPU, which translates to a 41.7% registration performance improvements and 3.0× power reduction.

show abstract

“…RRT performance depends mainly on efficiency of complex searches in the data structure optimised for representation of sparse data -Kd-Tree. FPGA solutions to Kd-tree searches were presented in [10], [11] although for different algorithms in image processing. RRT in UAV path planning operates on much smaller data sets and can be efficiently implemented as hardware-software solution on Zynq SoC, where FPGA fabric is tightly interconnected with a dual core processor.…”

Section: Online Motion Planning With Rrtmentioning

confidence: 99%

Development of control platform for micro-UAVs with integrated motion planning in wind

Doshi

Postula

Singh

et al. 2014

2014 3rd Mediterranean Conference on Embedded Computing (MECO)

View full text Add to dashboard Cite

Micro-UAVs (Unmanned Aerial Vehicles) attract much interest in a wide range of commercial applications. Due to small size, very limited power and susceptibility to wind their design poses difficult problems in control and motion planning. FPGA technology allows integration of high speed control circuits with efficient hardware implementations of complex algorithms on the same reconfigurable hardware chip. This is a very attractive option for implementation of a micro-UAV computing platform. This paper presents development of control platform for micro-UAVs focusing on integration of lower level control with online motion planning in wind. Some aspects of tight integration of on-board optical sensors with the aircraft control system are also discussed.

show abstract

An FPGA Acceleration for the Kd-tree Search in Photon Mapping

Cited by 11 publications

References 8 publications

Mesorasi: Architecture Support for Point Cloud Analytics via Delayed-Aggregation

Mesorasi: Architecture Support for Point Cloud Analytics via Delayed-Aggregation

Tigris

Development of control platform for micro-UAVs with integrated motion planning in wind

Contact Info

Product

Resources

About